Showing multiple columns in aggregate function including strings/characters in R - r

Question

R noob question here.
Let's say I have this data frame:
City State Pop
Fresno CA 494
San Franciso CA 805
San Jose CA 945
San Diego CA 1307
Los Angeles CA 3792
Reno NV 225
Henderson NV 257
Las Vegas NV 583
Gresham OR 105
Salem OR 154
Eugene OR 156
Portland OR 583
Fort Worth TX 741
Austin TX 790
Dallas TX 1197
San Antonio TX 1327
Houston TX 2100
I want to get let's say every 3rd lowest population per State, which would have:
City State Pop
San Jose CA 945
Las Vegas NV 583
Eugene OR 156
Dallas TX 1197
I tried this one:
ord_pop_state <- aggregate(Pop ~ State , data = ord_pop, function(x) { x[3] } )
And I get this one:
State Pop
CA 945
NV 583
OR 156
TX 1197
What do I lack on this one, in order for me to get the desired output that includes the City?

Answer 1

I would suggest to try data.table package for such task as the syntax is easier and the code is more efficient. I would also suggest to add order function in order to make sure that the data is sorted
library(data.table)
setDT(ord_pop)[order(Pop), .SD[3L], keyby = State]
# State City Pop
# 1: CA San Jose 945
# 2: NV Las Vegas 583
# 3: OR Eugene 156
# 4: TX Dallas 1197
So basically, first the data was ordered by Pop, then we subsetted .SD (which the notation parameter of the data itself) by State
Though this is easily solvable with base R too (we will assume that the data is sorted here), we can just create an index per group and then just do a simple subset by that index
ord_pop$indx <- with(ord_pop, ave(Pop, State, FUN = seq))
ord_pop[ord_pop$indx == 3L, ]
# City State Pop indx
# 3 San Jose CA 945 3
# 8 Las Vegas NV 583 3
# 11 Eugene OR 156 3
# 15 Dallas TX 1197 3

Answer 2

Here is a dplyr version:
df2 <- df %>%
group_by(state) %>% # Group observations by state
arrange(-pop) %>% # Within those groups, sort in descending order by pop
slice(3) # Extract the third row in each arranged group
Here's the toy data I used to test it:
set.seed(1)
df <- data.frame(state = rep(LETTERS[1:3], each = 5), city = rep(letters[1:5], 3), pop = round(rnorm(15, 1000, 100), digits=0))
And here's the output from that; it's a coincidence that 'b' was third-largest in each case, not a glitch in the code:
> df2
Source: local data frame [3 x 3]
Groups: state
state city pop
1 A b 1018
2 B b 1049
3 C b 1039

Answer 3

In R same end results can be achieved using different packages.Choice of package is a trade-off between efficiency and simplicity of code.
Since you come from a strong SQL background,this might be easier to use:
library(sqldf)
#Example to return 3rd lowest population of a State
result <-sqldf('Select City,State,Pop from data order by Pop limit 1 offset 2;')
#Note the SQL query is a sample and needs to be modifed to get desired result.